Elias Bareinboim
Columbia University, NY·Computer Science
More and more decisions are made by AI systems or with their help. Yet today's AI is often hard to explain, fragile and slow to adapt to new situations, which makes it hard to trust.
- Fields
Computer Science
Mathematics & Statistics
Engineering
- Topics
causal inference
causality
structural causal models
- Grants
- 1 known grant, through Sep 2028
What they work on
- Develop causality-integrated methods for online and offline policy learning, interventional planning, imitation learning and knowledge transfer so autonomous agents decide robustly with less data, building on structural causal models.
- Use causal models to help human-in-the-loop decision-support systems explain their choices, decide when to involve people, recognize their own limits and make fair, value-aligned decisions.
- Scale causal decision-making tools, balance trade-offs between explainability and decision quality, learn causal models of the world from data, and evaluate them on robotics and public health use-cases.
More about the research
The project argues that robust decisions require understanding cause and effect in the environment, and that most AI frameworks do not model causes explicitly. It aims to combine causal modeling with AI decision-making and learning. The goal is systems that need less data, explain their choices, adapt to change, and are safer.
Why it matters
AI systems increasingly make or support decisions in areas like robotics and public health. Systems that reason about cause and effect could need less data, adapt better when conditions change, and explain their choices to the people who rely on them. This could make automated decision support safer and fairer for its users.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Nan Jiang
University of Illinois Urbana-Champaign, IL·Computer Science·Early-career
Develop model selection methods for offline reinforcement learning that choose among candidate policies using a holdout dataset, building on theoretical work on value-function selection.
Kasim Candan
Arizona State University Campus Immersion, AZ·Computer Science
Develop PanAX methods that separate setting-specific from transferable parts of epidemic data and models, so knowledge from data-rich outbreaks can be reused for new outbreaks with limited data.
Faramarz Fekri
Georgia Institute of Technology, GA·Computer Science
Identify when cause-and-effect relationships can be calculated from observed data even when unmeasured confounders (hidden variables that affect both cause and outcome) are present, and develop robust ways to estimate those effects.
The school
Columbia University: 89 labs on PI Pages, $1.4B a year on research, 625 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.